The $1.6 Trillion AI Stampede: Is Wall Street Building Data Centers No One Will Use?

A century after Roger Babson warned of a looming crash ahead of the Great Depression, a similar tension is resurfacing as investors weigh the pace and scale of today's AI boom. Enthusiasm around artificial intelligence is echoing the optimism once attached to radios and automobiles in 1929, even as questions grow around valuation discipline and capital allocation. Tech companies are projected to spend just under $1.6 trillion a year on data centers by 2030, based on Omdia data, despite AI's long-term profit potential still being debated. Some observers suggest the issue may extend beyond AI itself, pointing to parallel signs of speculative behavior across markets, from sharply higher gold prices to record government deficits and the rapid expansion of a roughly $3 trillion private credit market increasingly tied to AI-related lending.Those speculative dynamics have already shown volatility elsewhere. Bitcoin's total market value climbed by $636 billion earlier in the year before reversing by mid-December, while memecoin trading volumes fell sharply from January peaks, according to Blockworks. In many cases, investors appeared to focus less on long-term value creation and more on the possibility of quick gains, a pattern reinforced by social media, global information flows, and survey data showing rising aspirations for extreme wealth, particularly among younger cohorts. Against this backdrop, AI has emerged as the dominant focal point in business and finance, potentially amplifying herd behavior as companies race to avoid being left behind.The pressure is now filtering into corporate balance sheets. While the largest technology firms are largely funding data center expansion from cash-rich positions, others are leaning more heavily on debt. Oracle is raising $38 billion to build data centers in Texas and Wisconsin, while newer operators such as CoreWeave and Fluidstack are also borrowing aggressively to finance AI-focused infrastructure. Some of these projects rely on complex financing structures to acquire Nvidia graphics processors, assets that some observers think could depreciate faster than expected if demand assumptions change. As investment firms build tens of billions of dollars' worth of capacity without fully secured customers, concerns are growing that a shift in sentiment could expose vulnerabilities reminiscent of past cycles, where optimism persisted until doubt finally took hold.